On Thu, Feb 09, 2023 at 12:18:55PM -0800, Andres Freund wrote:
> On 2023-02-09 11:39:17 -0800, Nathan Bossart wrote:
> Personally I'd probably squash these into a single commit.
done
> I'd probably mention that this should typically be of server lifetime / a
> 'static const' struct. Tableam documents this as follows:
done
>> + <note>
>> + <para>
>> + <varname>archive_library</varname> is only loaded in the archiver
>> process.
>> + </para>
>> + </note>
>> </sect1>
>
> That's not really related to any of the changes here, right?
>
> I'm not sure it's a good idea to document that. We e.g. probably should allow
> the library to check that the configuration is correct, at postmaster start,
> rather than later, at runtime.
removed
>> + <sect2 id="archive-module-startup">
>> + <title>Startup Callback</title>
>> + <para>
>> + The <function>startup_cb</function> callback is called shortly after the
>> + module is loaded. This callback can be used to perform any additional
>> + initialization required. If the archive module needs to have a state,
>> it
>> + can use <structfield>state->private_data</structfield> to store it.
>
> s/needs to have a state/has state/?
done
>> <programlisting>
>> -typedef bool (*ArchiveCheckConfiguredCB) (void);
>> +typedef bool (*ArchiveCheckConfiguredCB) (ArchiveModuleState *state);
>> </programlisting>
>>
>> If <literal>true</literal> is returned, the server will proceed with
>
> Hm. I wonder if ArchiveCheckConfiguredCB() should actually work without the
> state. We're not really doing anything yet, at that point, so it shouldn't
> really need state?
>
> The reason I'm wondering is that I think we should consider calling this from
> the GUC assignment hook, at least in postmaster. Which would make it more
> convenient to not have state, I think?
I agree that this callback should typically not need the state, but I'm not
sure whether it fits into the assignment hook for archive_library. This
callback is primarily meant for situations when you have archiving enabled,
but your module isn't configured yet (e.g., archive_command is empty). In
this case, we keep the WAL around, but we don't try to archive it until
this hook returns true. It's up to each module to define that criteria. I
can imagine someone introducing a GUC in their archive module that
temporarily halts archiving via this callback. In that case, calling it
via a GUC assignment hook probably won't work. In fact, checking whether
archive_command is empty in that hook might not work either.
>> <programlisting>
>> -typedef void (*ArchiveShutdownCB) (void);
>> +typedef void (*ArchiveShutdownCB) (ArchiveModuleState *state);
>> </programlisting>
>> </para>
>> </sect2>
>
> Perhaps mention that this needs to free state it allocated in the
> ArchiveModuleState, or it'll be leaked?
done
I left this out originally because the archiver exits shortly after calling
this. However, if you have DSM segments or something, it's probably wise
to make sure those are cleaned up. And I suppose we might not always exit
immediately after this callback, so establishing the habit of freeing the
state could be a good idea. In addition to updating this part of the docs,
I wrote a shutdown callback for basic_archive that frees its state.
--
Nathan Bossart
Amazon Web Services: https://aws.amazon.com
>From fab6305c344deb83c5b5c30ca2c09dbe1925a36c Mon Sep 17 00:00:00 2001
From: Nathan Bossart <nathandboss...@gmail.com>
Date: Thu, 9 Feb 2023 13:49:46 -0800
Subject: [PATCH v10 1/1] restructure archive modules API
---
contrib/basic_archive/basic_archive.c | 91 +++++++++++++++++++++----
doc/src/sgml/archive-modules.sgml | 35 +++++++---
src/backend/postmaster/pgarch.c | 27 +++++---
src/backend/postmaster/shell_archive.c | 34 +++++----
src/backend/utils/misc/guc_tables.c | 1 +
src/include/postmaster/archive_module.h | 58 ++++++++++++++++
src/include/postmaster/pgarch.h | 39 -----------
src/include/postmaster/shell_archive.h | 24 +++++++
8 files changed, 224 insertions(+), 85 deletions(-)
create mode 100644 src/include/postmaster/archive_module.h
create mode 100644 src/include/postmaster/shell_archive.h
diff --git a/contrib/basic_archive/basic_archive.c b/contrib/basic_archive/basic_archive.c
index 36b7a4814a..e4742c9c94 100644
--- a/contrib/basic_archive/basic_archive.c
+++ b/contrib/basic_archive/basic_archive.c
@@ -32,7 +32,7 @@
#include "common/int.h"
#include "miscadmin.h"
-#include "postmaster/pgarch.h"
+#include "postmaster/archive_module.h"
#include "storage/copydir.h"
#include "storage/fd.h"
#include "utils/guc.h"
@@ -40,14 +40,27 @@
PG_MODULE_MAGIC;
+typedef struct BasicArchiveData
+{
+ MemoryContext context;
+} BasicArchiveData;
+
static char *archive_directory = NULL;
-static MemoryContext basic_archive_context;
-static bool basic_archive_configured(void);
-static bool basic_archive_file(const char *file, const char *path);
+static void basic_archive_startup(ArchiveModuleState *state);
+static bool basic_archive_configured(ArchiveModuleState *state);
+static bool basic_archive_file(ArchiveModuleState *state, const char *file, const char *path);
static void basic_archive_file_internal(const char *file, const char *path);
static bool check_archive_directory(char **newval, void **extra, GucSource source);
static bool compare_files(const char *file1, const char *file2);
+static void basic_archive_shutdown(ArchiveModuleState *state);
+
+static const ArchiveModuleCallbacks basic_archive_callbacks = {
+ .startup_cb = basic_archive_startup,
+ .check_configured_cb = basic_archive_configured,
+ .archive_file_cb = basic_archive_file,
+ .shutdown_cb = basic_archive_shutdown
+};
/*
* _PG_init
@@ -67,10 +80,6 @@ _PG_init(void)
check_archive_directory, NULL, NULL);
MarkGUCPrefixReserved("basic_archive");
-
- basic_archive_context = AllocSetContextCreate(TopMemoryContext,
- "basic_archive",
- ALLOCSET_DEFAULT_SIZES);
}
/*
@@ -78,11 +87,28 @@ _PG_init(void)
*
* Returns the module's archiving callbacks.
*/
+const ArchiveModuleCallbacks *
+_PG_archive_module_init(void)
+{
+ return &basic_archive_callbacks;
+}
+
+/*
+ * basic_archive_startup
+ *
+ * Creates the module's memory context.
+ */
void
-_PG_archive_module_init(ArchiveModuleCallbacks *cb)
+basic_archive_startup(ArchiveModuleState *state)
{
- cb->check_configured_cb = basic_archive_configured;
- cb->archive_file_cb = basic_archive_file;
+ BasicArchiveData *data;
+
+ data = (BasicArchiveData *) MemoryContextAllocZero(TopMemoryContext,
+ sizeof(BasicArchiveData));
+ data->context = AllocSetContextCreate(TopMemoryContext,
+ "basic_archive",
+ ALLOCSET_DEFAULT_SIZES);
+ state->private_data = (void *) data;
}
/*
@@ -133,7 +159,7 @@ check_archive_directory(char **newval, void **extra, GucSource source)
* Checks that archive_directory is not blank.
*/
static bool
-basic_archive_configured(void)
+basic_archive_configured(ArchiveModuleState *state)
{
return archive_directory != NULL && archive_directory[0] != '\0';
}
@@ -144,10 +170,12 @@ basic_archive_configured(void)
* Archives one file.
*/
static bool
-basic_archive_file(const char *file, const char *path)
+basic_archive_file(ArchiveModuleState *state, const char *file, const char *path)
{
sigjmp_buf local_sigjmp_buf;
MemoryContext oldcontext;
+ BasicArchiveData *data = (BasicArchiveData *) (state->private_data);
+ MemoryContext basic_archive_context = data->context;
/*
* We run basic_archive_file_internal() in our own memory context so that
@@ -366,3 +394,40 @@ compare_files(const char *file1, const char *file2)
return ret;
}
+
+/*
+ * basic_archive_shutdown
+ *
+ * Frees our allocated state.
+ */
+static void
+basic_archive_shutdown(ArchiveModuleState *state)
+{
+ BasicArchiveData *data;
+ MemoryContext basic_archive_context;
+
+ /*
+ * If we didn't get to storing the pointer to our allocated state, we don't
+ * have anything to clean up.
+ */
+ data = (BasicArchiveData *) (state->private_data);
+ if (data == NULL)
+ return;
+
+ /*
+ * If called within basic_archive_context, switch out of it before we
+ * delete it.
+ */
+ basic_archive_context = data->context;
+ if (CurrentMemoryContext == basic_archive_context)
+ MemoryContextSwitchTo(TopMemoryContext);
+
+ if (MemoryContextIsValid(basic_archive_context))
+ MemoryContextDelete(basic_archive_context);
+
+ /*
+ * Finally, free the state.
+ */
+ pfree(data);
+ state->private_data = NULL;
+}
diff --git a/doc/src/sgml/archive-modules.sgml b/doc/src/sgml/archive-modules.sgml
index ef02051f7f..668eb34991 100644
--- a/doc/src/sgml/archive-modules.sgml
+++ b/doc/src/sgml/archive-modules.sgml
@@ -47,18 +47,22 @@
normal library search path is used to locate the library. To provide the
required archive module callbacks and to indicate that the library is
actually an archive module, it needs to provide a function named
- <function>_PG_archive_module_init</function>. This function is passed a
- struct that needs to be filled with the callback function pointers for
- individual actions.
+ <function>_PG_archive_module_init</function>. The result of the function
+ must be a pointer to a struct of type
+ <structname>ArchiveModuleCallbacks</structname>, which contains everything
+ that the core code needs to know how to make use of the archive module. The
+ return value needs to be of server lifetime, which is typically achieved by
+ defining it as a <literal>static const</literal> variable in global scope.
<programlisting>
typedef struct ArchiveModuleCallbacks
{
+ ArchiveStartupCB startup_cb;
ArchiveCheckConfiguredCB check_configured_cb;
ArchiveFileCB archive_file_cb;
ArchiveShutdownCB shutdown_cb;
} ArchiveModuleCallbacks;
-typedef void (*ArchiveModuleInit) (struct ArchiveModuleCallbacks *cb);
+typedef const ArchiveModuleCallbacks *(*ArchiveModuleInit) (void);
</programlisting>
Only the <function>archive_file_cb</function> callback is required. The
@@ -73,6 +77,20 @@ typedef void (*ArchiveModuleInit) (struct ArchiveModuleCallbacks *cb);
The server will call them as required to process each individual WAL file.
</para>
+ <sect2 id="archive-module-startup">
+ <title>Startup Callback</title>
+ <para>
+ The <function>startup_cb</function> callback is called shortly after the
+ module is loaded. This callback can be used to perform any additional
+ initialization required. If the archive module has state, it can use
+ <structfield>state->private_data</structfield> to store it.
+
+<programlisting>
+typedef void (*ArchiveStartupCB) (ArchiveModuleState *state);
+</programlisting>
+ </para>
+ </sect2>
+
<sect2 id="archive-module-check">
<title>Check Callback</title>
<para>
@@ -83,7 +101,7 @@ typedef void (*ArchiveModuleInit) (struct ArchiveModuleCallbacks *cb);
assumes the module is configured.
<programlisting>
-typedef bool (*ArchiveCheckConfiguredCB) (void);
+typedef bool (*ArchiveCheckConfiguredCB) (ArchiveModuleState *state);
</programlisting>
If <literal>true</literal> is returned, the server will proceed with
@@ -105,7 +123,7 @@ WARNING: archive_mode enabled, yet archiving is not configured
single WAL file.
<programlisting>
-typedef bool (*ArchiveFileCB) (const char *file, const char *path);
+typedef bool (*ArchiveFileCB) (ArchiveModuleState *state, const char *file, const char *path);
</programlisting>
If <literal>true</literal> is returned, the server proceeds as if the file
@@ -125,10 +143,11 @@ typedef bool (*ArchiveFileCB) (const char *file, const char *path);
process exits (e.g., after an error) or the value of
<xref linkend="guc-archive-library"/> changes. If no
<function>shutdown_cb</function> is defined, no special action is taken in
- these situations.
+ these situations. If the archive module has state, this callback should
+ free it to avoid leaks.
<programlisting>
-typedef void (*ArchiveShutdownCB) (void);
+typedef void (*ArchiveShutdownCB) (ArchiveModuleState *state);
</programlisting>
</para>
</sect2>
diff --git a/src/backend/postmaster/pgarch.c b/src/backend/postmaster/pgarch.c
index e551af2905..99a07996f7 100644
--- a/src/backend/postmaster/pgarch.c
+++ b/src/backend/postmaster/pgarch.c
@@ -34,8 +34,10 @@
#include "lib/binaryheap.h"
#include "libpq/pqsignal.h"
#include "pgstat.h"
+#include "postmaster/archive_module.h"
#include "postmaster/interrupt.h"
#include "postmaster/pgarch.h"
+#include "postmaster/shell_archive.h"
#include "storage/fd.h"
#include "storage/ipc.h"
#include "storage/latch.h"
@@ -97,7 +99,8 @@ char *XLogArchiveLibrary = "";
*/
static time_t last_sigterm_time = 0;
static PgArchData *PgArch = NULL;
-static ArchiveModuleCallbacks ArchiveContext;
+static const ArchiveModuleCallbacks *ArchiveCallbacks;
+static ArchiveModuleState *archive_module_state;
/*
@@ -406,8 +409,8 @@ pgarch_ArchiverCopyLoop(void)
HandlePgArchInterrupts();
/* can't do anything if not configured ... */
- if (ArchiveContext.check_configured_cb != NULL &&
- !ArchiveContext.check_configured_cb())
+ if (ArchiveCallbacks->check_configured_cb != NULL &&
+ !ArchiveCallbacks->check_configured_cb(archive_module_state))
{
ereport(WARNING,
(errmsg("archive_mode enabled, yet archiving is not configured")));
@@ -508,7 +511,7 @@ pgarch_archiveXlog(char *xlog)
snprintf(activitymsg, sizeof(activitymsg), "archiving %s", xlog);
set_ps_display(activitymsg);
- ret = ArchiveContext.archive_file_cb(xlog, pathname);
+ ret = ArchiveCallbacks->archive_file_cb(archive_module_state, xlog, pathname);
if (ret)
snprintf(activitymsg, sizeof(activitymsg), "last was %s", xlog);
else
@@ -814,7 +817,7 @@ HandlePgArchInterrupts(void)
/*
* LoadArchiveLibrary
*
- * Loads the archiving callbacks into our local ArchiveContext.
+ * Loads the archiving callbacks into our local ArchiveCallbacks.
*/
static void
LoadArchiveLibrary(void)
@@ -827,8 +830,6 @@ LoadArchiveLibrary(void)
errmsg("both archive_command and archive_library set"),
errdetail("Only one of archive_command, archive_library may be set.")));
- memset(&ArchiveContext, 0, sizeof(ArchiveModuleCallbacks));
-
/*
* If shell archiving is enabled, use our special initialization function.
* Otherwise, load the library and call its _PG_archive_module_init().
@@ -844,12 +845,16 @@ LoadArchiveLibrary(void)
ereport(ERROR,
(errmsg("archive modules have to define the symbol %s", "_PG_archive_module_init")));
- (*archive_init) (&ArchiveContext);
+ ArchiveCallbacks = (*archive_init) ();
- if (ArchiveContext.archive_file_cb == NULL)
+ if (ArchiveCallbacks->archive_file_cb == NULL)
ereport(ERROR,
(errmsg("archive modules must register an archive callback")));
+ archive_module_state = (ArchiveModuleState *) palloc0(sizeof(ArchiveModuleState));
+ if (ArchiveCallbacks->startup_cb != NULL)
+ ArchiveCallbacks->startup_cb(archive_module_state);
+
before_shmem_exit(pgarch_call_module_shutdown_cb, 0);
}
@@ -859,6 +864,6 @@ LoadArchiveLibrary(void)
static void
pgarch_call_module_shutdown_cb(int code, Datum arg)
{
- if (ArchiveContext.shutdown_cb != NULL)
- ArchiveContext.shutdown_cb();
+ if (ArchiveCallbacks->shutdown_cb != NULL)
+ ArchiveCallbacks->shutdown_cb(archive_module_state);
}
diff --git a/src/backend/postmaster/shell_archive.c b/src/backend/postmaster/shell_archive.c
index 7771b951b7..54b02ca152 100644
--- a/src/backend/postmaster/shell_archive.c
+++ b/src/backend/postmaster/shell_archive.c
@@ -20,28 +20,34 @@
#include "access/xlog.h"
#include "common/percentrepl.h"
#include "pgstat.h"
-#include "postmaster/pgarch.h"
-
-static bool shell_archive_configured(void);
-static bool shell_archive_file(const char *file, const char *path);
-static void shell_archive_shutdown(void);
-
-void
-shell_archive_init(ArchiveModuleCallbacks *cb)
+#include "postmaster/archive_module.h"
+#include "postmaster/shell_archive.h"
+
+static bool shell_archive_configured(ArchiveModuleState *state);
+static bool shell_archive_file(ArchiveModuleState *state, const char *file, const char *path);
+static void shell_archive_shutdown(ArchiveModuleState *state);
+
+static const ArchiveModuleCallbacks shell_archive_callbacks = {
+ .startup_cb = NULL,
+ .check_configured_cb = shell_archive_configured,
+ .archive_file_cb = shell_archive_file,
+ .shutdown_cb = shell_archive_shutdown
+};
+
+const ArchiveModuleCallbacks *
+shell_archive_init(void)
{
- cb->check_configured_cb = shell_archive_configured;
- cb->archive_file_cb = shell_archive_file;
- cb->shutdown_cb = shell_archive_shutdown;
+ return &shell_archive_callbacks;
}
static bool
-shell_archive_configured(void)
+shell_archive_configured(ArchiveModuleState *state)
{
return XLogArchiveCommand[0] != '\0';
}
static bool
-shell_archive_file(const char *file, const char *path)
+shell_archive_file(ArchiveModuleState *state, const char *file, const char *path)
{
char *xlogarchcmd;
char *nativePath = NULL;
@@ -123,7 +129,7 @@ shell_archive_file(const char *file, const char *path)
}
static void
-shell_archive_shutdown(void)
+shell_archive_shutdown(ArchiveModuleState *state)
{
elog(DEBUG1, "archiver process shutting down");
}
diff --git a/src/backend/utils/misc/guc_tables.c b/src/backend/utils/misc/guc_tables.c
index b46e3b8c55..c3ee83b691 100644
--- a/src/backend/utils/misc/guc_tables.c
+++ b/src/backend/utils/misc/guc_tables.c
@@ -52,6 +52,7 @@
#include "parser/parse_expr.h"
#include "parser/parser.h"
#include "pgstat.h"
+#include "postmaster/archive_module.h"
#include "postmaster/autovacuum.h"
#include "postmaster/bgworker_internals.h"
#include "postmaster/bgwriter.h"
diff --git a/src/include/postmaster/archive_module.h b/src/include/postmaster/archive_module.h
new file mode 100644
index 0000000000..79e9d53534
--- /dev/null
+++ b/src/include/postmaster/archive_module.h
@@ -0,0 +1,58 @@
+/*-------------------------------------------------------------------------
+ *
+ * archive_module.h
+ * Exports for archive modules.
+ *
+ * Copyright (c) 2022-2023, PostgreSQL Global Development Group
+ *
+ * src/include/postmaster/archive_module.h
+ *
+ *-------------------------------------------------------------------------
+ */
+#ifndef _ARCHIVE_MODULE_H
+#define _ARCHIVE_MODULE_H
+
+/*
+ * The value of the archive_library GUC.
+ */
+extern PGDLLIMPORT char *XLogArchiveLibrary;
+
+typedef struct ArchiveModuleState
+{
+ /*
+ * Private data pointer for use by an archive module. This can be used to
+ * store state for the module that will be passed to each of its callbacks.
+ */
+ void *private_data;
+} ArchiveModuleState;
+
+/*
+ * Archive module callbacks
+ *
+ * These callback functions should be defined by archive libraries and returned
+ * via _PG_archive_module_init(). ArchiveFileCB is the only required callback.
+ * For more information about the purpose of each callback, refer to the
+ * archive modules documentation.
+ */
+typedef void (*ArchiveStartupCB) (ArchiveModuleState *state);
+typedef bool (*ArchiveCheckConfiguredCB) (ArchiveModuleState *state);
+typedef bool (*ArchiveFileCB) (ArchiveModuleState *state, const char *file, const char *path);
+typedef void (*ArchiveShutdownCB) (ArchiveModuleState *state);
+
+typedef struct ArchiveModuleCallbacks
+{
+ ArchiveStartupCB startup_cb;
+ ArchiveCheckConfiguredCB check_configured_cb;
+ ArchiveFileCB archive_file_cb;
+ ArchiveShutdownCB shutdown_cb;
+} ArchiveModuleCallbacks;
+
+/*
+ * Type of the shared library symbol _PG_archive_module_init that is looked
+ * up when loading an archive library.
+ */
+typedef const ArchiveModuleCallbacks *(*ArchiveModuleInit) (void);
+
+extern PGDLLEXPORT const ArchiveModuleCallbacks *_PG_archive_module_init(void);
+
+#endif /* _ARCHIVE_MODULE_H */
diff --git a/src/include/postmaster/pgarch.h b/src/include/postmaster/pgarch.h
index bcd51dfad6..3bd4fac71e 100644
--- a/src/include/postmaster/pgarch.h
+++ b/src/include/postmaster/pgarch.h
@@ -33,43 +33,4 @@ extern void PgArchiverMain(void) pg_attribute_noreturn();
extern void PgArchWakeup(void);
extern void PgArchForceDirScan(void);
-/*
- * The value of the archive_library GUC.
- */
-extern PGDLLIMPORT char *XLogArchiveLibrary;
-
-/*
- * Archive module callbacks
- *
- * These callback functions should be defined by archive libraries and returned
- * via _PG_archive_module_init(). ArchiveFileCB is the only required callback.
- * For more information about the purpose of each callback, refer to the
- * archive modules documentation.
- */
-typedef bool (*ArchiveCheckConfiguredCB) (void);
-typedef bool (*ArchiveFileCB) (const char *file, const char *path);
-typedef void (*ArchiveShutdownCB) (void);
-
-typedef struct ArchiveModuleCallbacks
-{
- ArchiveCheckConfiguredCB check_configured_cb;
- ArchiveFileCB archive_file_cb;
- ArchiveShutdownCB shutdown_cb;
-} ArchiveModuleCallbacks;
-
-/*
- * Type of the shared library symbol _PG_archive_module_init that is looked
- * up when loading an archive library.
- */
-typedef void (*ArchiveModuleInit) (ArchiveModuleCallbacks *cb);
-
-extern PGDLLEXPORT void _PG_archive_module_init(ArchiveModuleCallbacks *cb);
-
-/*
- * Since the logic for archiving via a shell command is in the core server
- * and does not need to be loaded via a shared library, it has a special
- * initialization function.
- */
-extern void shell_archive_init(ArchiveModuleCallbacks *cb);
-
#endif /* _PGARCH_H */
diff --git a/src/include/postmaster/shell_archive.h b/src/include/postmaster/shell_archive.h
new file mode 100644
index 0000000000..be83c7715a
--- /dev/null
+++ b/src/include/postmaster/shell_archive.h
@@ -0,0 +1,24 @@
+/*-------------------------------------------------------------------------
+ *
+ * shell_archive.h
+ * Exports for archiving via shell.
+ *
+ * Copyright (c) 2022-2023, PostgreSQL Global Development Group
+ *
+ * src/include/postmaster/shell_archive.h
+ *
+ *-------------------------------------------------------------------------
+ */
+#ifndef _SHELL_ARCHIVE_H
+#define _SHELL_ARCHIVE_H
+
+#include "postmaster/archive_module.h"
+
+/*
+ * Since the logic for archiving via a shell command is in the core server
+ * and does not need to be loaded via a shared library, it has a special
+ * initialization function.
+ */
+extern const ArchiveModuleCallbacks *shell_archive_init(void);
+
+#endif /* _SHELL_ARCHIVE_H */
--
2.25.1
